神经网络解决多分类问题例：数字识别
目录
神经网络解决多分类问题例：数字识别
1. 观察样本(Visualizing the data)
训练集提供5000张数字图片，每张图片为20x20像素，并被转化成1x400的向量存储。样本输入为5000x400的矩阵，输出为5000x1的向量。coursera提供了将灰度值转化为图片的函数，但这对我们解决问题没有实质性的帮助。
2. 设计神经网络(Designing Nural Network)
由于每一个样本输入为1x400的向量，因此输入神经元应有400个。我们预测的数字共有10个，因此输出神经元应有10个。由于问题并不复杂，只需使用1隐层即可（早期的自动驾驶试验所使用的神经网络为3隐层），隐层中的神经元个数定为25个。
定义三个变量存储各层的神经元个数，方便后续调用。
input_layer_size = 400;
hidden_layer_size = 25;
output_layer_size = 10;
3. 编写代价函数计算函数(nnCostFunction)
3.1 实现必备的工具函数
在BP神经网络的实现过程中，有几个操作是经常使用的。如果不将其封装成函数很容易写着写着就把自己绕晕。
-
sigmoid函数
function g = sigmoid(z) g = 1 ./ (1+exp(-z)); end -
addBias函数
由于我们在设计神经网络时没有考虑偏置神经元，而在前向传播的时候，计算下一层的输入值必须用到偏置神经元的参数\(\theta_0\)，因此在这里封装成函数。
function ans = addBias(X) ans = [ones(size(X,1)),X]; end -
oneHot函数
one-hot-encoding译作独热编码。它用于将m*1的样本输出y转化为m*output_layer_size的矩阵。对于每一行，有output_layer_size个数，分别对应各个输出神经元是0还是1。
function ans = oneHot(y,output_layer_size) ans = zeros(size(y,1),output_layer_size); for i = 1:size(y,1) ans(i,y(i)) = 1; end end
3.2 规范化矩阵定义
在神经网络的实现过程中，矩阵维度的错误可能是初学者遇见最频繁的一个问题。而每一次发现矩阵维度错误时，往往又需要从输入层开始一步步推导出正确的矩阵维度，重新修改代码中矩阵的计算方式、计算顺序。出现这种情况，往往是编写代码时逻辑混乱，一下子把这个矩阵写出m*n,一下子把那个矩阵写成n*m。如果能够在编写代码前事先设计好各个矩阵的表示方法，规范化行、列的实际意义，就能在发现错误后快速改正，或者直接避免错误。
在这里我们沿用coursera的讲义中的矩阵定义的规范
| 矩阵名称 | 常用代号 | 矩阵规模 | 行意义 | 列意义 |
|---|---|---|---|---|
| 输入值矩阵 | \(X,z^{(i)}\) | m * n | 样本个数 | 该层神经元个数（属性个数） |
| 参数矩阵 | \(\Theta_{(j)}\) | (a+1) * b | 该层神经元个数（这一层的激活函数的个数） | 上一层神经元个数+1（每个激活函数的属性个数+偏置属性） |
| 输出值矩阵 | \(Y,a^{(i)}\) | m * k | 样本个数 | 该层神经元个数（输出值个数） |
另外简记: input_layer_size = n, hidden_layer_size = l, output_layer_size = K
3.3 实现前向传播(Feedforward )
前向传播是计算各个输出神经元的输出值的过程。它可以向量化计算。
将神经网络的前向传播定义在 nnCostFunction 函数中
传入参数[ nn_params(所有参数值\(\theta\)的展开向量) ， input_layer_size, hidden_layer_size, output_layer_size , X, y, lambda ]
function [J grad] = nnCostFunction(nn_params, ...
input_layer_size, ...
hidden_layer_size, ...
output_layer_size, ...
X, y, lambda)
% 使用reshape函数将向量nn_params重新构造成Theta1,Theta2两个矩阵。注意，Theta1，Theta2两个矩阵
% 都是考虑了偏置神经元的。
Theta1 = reshape(nn_params(1:hidden_layer_size * (input_layer_size + 1)), ...
hidden_layer_size, (input_layer_size + 1));
Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):end), ...
output_layer_size, (hidden_layer_size + 1));
% 样本个数
m = size(X, 1);
% 代价值
J = 0;
% 梯度值
Theta1_grad = zeros(size(Theta1));
Theta2_grad = zeros(size(Theta2));
% 初始化完成
% ==============================================================================
% ==============================================================================
% Part1: 实现前向传播
% 前向传播
z2 = addBias(X) * Theta1';
a2 = sigmoid(z2);
z3 = addBias(a2) * Theta2';
a3 = sigmoid(z3);
% 独热编码
encodeY = oneHot(y,output_layer_size);
% tempTheta 用于计算正则项，将偏置神经元对应的theta全部置0
tempTheta2 = Theta2;
tempTheta2(:,1) = 0;
tempTheta1 = Theta1;
tempTheta1(:,1) = 0;
J = 1/m * sum(sum((-encodeY .* log(a3) - (1-encodeY) .* log(1-a3)))) + 1/(2*m) * lambda * (sum(sum(tempTheta1 .^2)) + sum(sum(tempTheta2 .^ 2)) );
代价函数计算公式
\[J(\theta) = \frac{1}{m}\sum_{i=1}^{m}\sum_{k=1}^{K}[-y_k^{(i)}\log{(h_\theta(x^{(i)})_k)} - (1-y_k^{(i)})\log{(1-h_\theta(x^{(i)})_k)}] + \frac{\lambda}{2m}[\sum_{i=1}^l\sum_{j=1}^n\Theta_{ij}^2 + \sum_{i=1}^K\sum_{j=1}^l\Theta_{ij}^2]
\]
神经网络的学习能力过强，可以拟合高度复杂的模型。因此如果不加以正则化很可能会过拟合。此时经验误差很小而泛化误差不够小
3.4 实现反向传播(Backpropagation)
这一节是反向传播算法的核心部分，也是最难懂、写代码时最复杂的一部分。初学时肯定很头疼，我在这部分卡了3天没能理解。
这一部分设计链式法则和矩阵求导，其中链式法则必须会，矩阵求导如果不会可以先通过矩阵的维度来推导结果（推导得相对慢一些）。
推荐b站一个视频，详细的讲了反向传播算法的梯度值计算。她还发过吴恩达神经网络解数字识别的Python实现，我没看不过总体思路肯定和matlab实现类似，可以作为理解反向传播算法的一个讲解视频。
反向传播算法用于解决梯度计算。通过链式法则与矩阵求推导出各个参数\(\theta\)的梯度
Theta2_grad = 1/m*(a3-encodeY)' * addBias(a2) + lambda * tempTheta2/m;
Theta1_grad = 1/m*(Theta2(:,2:end)' * (a3-encodeY)' .* a2' .* (1-a2') * addBias(X)) + lambda * tempTheta1 / m;
% Unroll gradients
grad = [Theta1_grad(:) ; Theta2_grad(:)];
4. 高级最优化训练神经网络(Learning parameters using fmincg)
4.1 随机初始化参数
function W = randInitializeWeights(L_in, L_out)
W = zeros(L_out, 1 + L_in);
% Randomly initialize the weights to small values
epsilon_init = 0.12;
W = rand(L_out, 1 + L_in) * 2 * epsilon_init - epsilon_init;
end
随机初始化参数对于神经网络的重要性不必多提，属于基础知识，不明白的可以重新看一看吴恩达的那节课。
initial_Theta1 = randInitializeWeights(input_layer_size, hidden_layer_size);
initial_Theta2 = randInitializeWeights(hidden_layer_size, output_layer_size);
% 展开成向量，便于传递参数
initial_nn_params = [initial_Theta1(:) ; initial_Theta2(:)];
4.2 调用高级最优化函数训练神经网络
这是coursera给出的高级最优化函数 fmincg。使用前先学习用法，如果不会用就用fminunc，略慢一些。
function [X, fX, i] = fmincg(f, X, options, P1, P2, P3, P4, P5)
% Minimize a continuous differentialble multivariate function. Starting point
% is given by "X" (D by 1), and the function named in the string "f", must
% return a function value and a vector of partial derivatives. The Polack-
% Ribiere flavour of conjugate gradients is used to compute search directions,
% and a line search using quadratic and cubic polynomial approximations and the
% Wolfe-Powell stopping criteria is used together with the slope ratio method
% for guessing initial step sizes. Additionally a bunch of checks are made to
% make sure that exploration is taking place and that extrapolation will not
% be unboundedly large. The "length" gives the length of the run: if it is
% positive, it gives the maximum number of line searches, if negative its
% absolute gives the maximum allowed number of function evaluations. You can
% (optionally) give "length" a second component, which will indicate the
% reduction in function value to be expected in the first line-search (defaults
% to 1.0). The function returns when either its length is up, or if no further
% progress can be made (ie, we are at a minimum, or so close that due to
% numerical problems, we cannot get any closer). If the function terminates
% within a few iterations, it could be an indication that the function value
% and derivatives are not consistent (ie, there may be a bug in the
% implementation of your "f" function). The function returns the found
% solution "X", a vector of function values "fX" indicating the progress made
% and "i" the number of iterations (line searches or function evaluations,
% depending on the sign of "length") used.
%
% Usage: [X, fX, i] = fmincg(f, X, options, P1, P2, P3, P4, P5)
%
% See also: checkgrad
%
% Copyright (C) 2001 and 2002 by Carl Edward Rasmussen. Date 2002-02-13
%
%
% (C) Copyright 1999, 2000 & 2001, Carl Edward Rasmussen
%
% Permission is granted for anyone to copy, use, or modify these
% programs and accompanying documents for purposes of research or
% education, provided this copyright notice is retained, and note is
% made of any changes that have been made.
%
% These programs and documents are distributed without any warranty,
% express or implied. As the programs were written for research
% purposes only, they have not been tested to the degree that would be
% advisable in any important application. All use of these programs is
% entirely at the user's own risk.
%
% [ml-class] Changes Made:
% 1) Function name and argument specifications
% 2) Output display
%
% Read options
if exist('options', 'var') && ~isempty(options) && isfield(options, 'MaxIter')
length = options.MaxIter;
else
length = 100;
end
RHO = 0.01; % a bunch of constants for line searches
SIG = 0.5; % RHO and SIG are the constants in the Wolfe-Powell conditions
INT = 0.1; % don't reevaluate within 0.1 of the limit of the current bracket
EXT = 3.0; % extrapolate maximum 3 times the current bracket
MAX = 20; % max 20 function evaluations per line search
RATIO = 100; % maximum allowed slope ratio
argstr = ['feval(f, X']; % compose string used to call function
for i = 1:(nargin - 3)
argstr = [argstr, ',P', int2str(i)];
end
argstr = [argstr, ')'];
if max(size(length)) == 2, red=length(2); length=length(1); else red=1; end
S=['Iteration '];
i = 0; % zero the run length counter
ls_failed = 0; % no previous line search has failed
fX = [];
[f1 df1] = eval(argstr); % get function value and gradient
i = i + (length<0); % count epochs?!
s = -df1; % search direction is steepest
d1 = -s'*s; % this is the slope
z1 = red/(1-d1); % initial step is red/(|s|+1)
while i < abs(length) % while not finished
i = i + (length>0); % count iterations?!
X0 = X; f0 = f1; df0 = df1; % make a copy of current values
X = X + z1*s; % begin line search
[f2 df2] = eval(argstr);
i = i + (length<0); % count epochs?!
d2 = df2'*s;
f3 = f1; d3 = d1; z3 = -z1; % initialize point 3 equal to point 1
if length>0, M = MAX; else M = min(MAX, -length-i); end
success = 0; limit = -1; % initialize quanteties
while 1
while ((f2 > f1+z1*RHO*d1) || (d2 > -SIG*d1)) && (M > 0)
limit = z1; % tighten the bracket
if f2 > f1
z2 = z3 - (0.5*d3*z3*z3)/(d3*z3+f2-f3); % quadratic fit
else
A = 6*(f2-f3)/z3+3*(d2+d3); % cubic fit
B = 3*(f3-f2)-z3*(d3+2*d2);
z2 = (sqrt(B*B-A*d2*z3*z3)-B)/A; % numerical error possible - ok!
end
if isnan(z2) || isinf(z2)
z2 = z3/2; % if we had a numerical problem then bisect
end
z2 = max(min(z2, INT*z3),(1-INT)*z3); % don't accept too close to limits
z1 = z1 + z2; % update the step
X = X + z2*s;
[f2 df2] = eval(argstr);
M = M - 1; i = i + (length<0); % count epochs?!
d2 = df2'*s;
z3 = z3-z2; % z3 is now relative to the location of z2
end
if f2 > f1+z1*RHO*d1 || d2 > -SIG*d1
break; % this is a failure
elseif d2 > SIG*d1
success = 1; break; % success
elseif M == 0
break; % failure
end
A = 6*(f2-f3)/z3+3*(d2+d3); % make cubic extrapolation
B = 3*(f3-f2)-z3*(d3+2*d2);
z2 = -d2*z3*z3/(B+sqrt(B*B-A*d2*z3*z3)); % num. error possible - ok!
if ~isreal(z2) || isnan(z2) || isinf(z2) || z2 < 0 % num prob or wrong sign?
if limit < -0.5 % if we have no upper limit
z2 = z1 * (EXT-1); % the extrapolate the maximum amount
else
z2 = (limit-z1)/2; % otherwise bisect
end
elseif (limit > -0.5) && (z2+z1 > limit) % extraplation beyond max?
z2 = (limit-z1)/2; % bisect
elseif (limit < -0.5) && (z2+z1 > z1*EXT) % extrapolation beyond limit
z2 = z1*(EXT-1.0); % set to extrapolation limit
elseif z2 < -z3*INT
z2 = -z3*INT;
elseif (limit > -0.5) && (z2 < (limit-z1)*(1.0-INT)) % too close to limit?
z2 = (limit-z1)*(1.0-INT);
end
f3 = f2; d3 = d2; z3 = -z2; % set point 3 equal to point 2
z1 = z1 + z2; X = X + z2*s; % update current estimates
[f2 df2] = eval(argstr);
M = M - 1; i = i + (length<0); % count epochs?!
d2 = df2'*s;
end % end of line search
if success % if line search succeeded
f1 = f2; fX = [fX' f1]';
fprintf('%s %4i | Cost: %4.6e\r', S, i, f1);
s = (df2'*df2-df1'*df2)/(df1'*df1)*s - df2; % Polack-Ribiere direction
tmp = df1; df1 = df2; df2 = tmp; % swap derivatives
d2 = df1'*s;
if d2 > 0 % new slope must be negative
s = -df1; % otherwise use steepest direction
d2 = -s'*s;
end
z1 = z1 * min(RATIO, d1/(d2-realmin)); % slope ratio but max RATIO
d1 = d2;
ls_failed = 0; % this line search did not fail
else
X = X0; f1 = f0; df1 = df0; % restore point from before failed line search
if ls_failed || i > abs(length) % line search failed twice in a row
break; % or we ran out of time, so we give up
end
tmp = df1; df1 = df2; df2 = tmp; % swap derivatives
s = -df1; % try steepest
d1 = -s'*s;
z1 = 1/(1-d1);
ls_failed = 1; % this line search failed
end
if exist('OCTAVE_VERSION')
fflush(stdout);
end
end
fprintf('\n');
调用部分
options = optimset('MaxIter', 50);
% You should also try different values of lambda
lambda = 1;
% Create "short hand" for the cost function to be minimized
costFunction = @(p) nnCostFunction(p, ...
input_layer_size, ...
hidden_layer_size, ...
output_layer_size, X, y, lambda);
% Now, costFunction is a function that takes in only one argument (the
% neural network parameters)
[nn_params, cost] = fmincg(costFunction, initial_nn_params, options);
% Obtain Theta1 and Theta2 back from nn_params
Theta1 = reshape(nn_params(1:hidden_layer_size * (input_layer_size + 1)), ...
hidden_layer_size, (input_layer_size + 1));
Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):end), ...
output_layer_size, (hidden_layer_size + 1));
5.计算经验误差 / 调整参数 / 模型评估(Trying out different learning settings)
function p = predict(Theta1, Theta2, X)
%PREDICT Predict the label of an input given a trained neural network
% p = PREDICT(Theta1, Theta2, X) outputs the predicted label of X given the
% trained weights of a neural network (Theta1, Theta2)
% Useful values
m = size(X, 1);
num_labels = size(Theta2, 1);
% You need to return the following variables correctly
p = zeros(size(X, 1), 1);
h1 = sigmoid([ones(m, 1) X] * Theta1');
h2 = sigmoid([ones(m, 1) h1] * Theta2');
[dummy, p] = max(h2, [], 2);
% =========================================================================
end
调用函数
pred = predict(Theta1, Theta2, X);
fprintf('\nTraining Set Accuracy: %f\n', mean(double(pred == y)) * 100);
调整参数，可以观察经验误差的变化
| \(\lambda\) | \(Max \ \ Iterations\) | \(accuracy\) |
|---|---|---|
| 1 | 50 | 94.34% ~ 96.00% |
| 1 | 100 | 98.64% |
| 1.5 | 1000 | 99.04% |
| 1.5 | 5000(折磨电脑) | 99.26% |
| 2 | 2000 | 98.68% |
这些都是经验误差，而我们的目的是减小泛化误差。然而coursera没有给出测试集，我们暂时不去测泛化误差了。严格来说还应该有泛化误差的测量部分。
6. 观察隐层神经元(Visualizing the hidden layer)
这一部分显得没有那么重要，我们可以看看隐层神经元到底在干些什么。
对于每一个隐层神经元，找到一组输入vector[1,400]使之激活值接近1（此时表示其极大可能性为某一种状态，而此时其余神经元均接近于0）。然后将其转化成20x20像素图像。
fprintf('\nVisualizing Neural Network... \n')
displayData(Theta1(:, 2:end));
fprintf('\nProgram paused. Press enter to continue.\n');
pause;
---- suffer now and live the rest of your life as a champion ----
